Machine learning detects early rising behavioral changes in mouse models of Parkinson’s disease

Detecting early rising Parkinson's disease (PD) symptoms could improve treatment outcomes by enabling earlier treatment interventions. In a new eNeuro paper, Daniil Berezhnoi, from Georgetown University, and colleagues used machine learning technology to detect subtle, early rising behavioral changes in mouse models of PD. The researchers also evaluated whether Levodopa, the primary approved treatment for PD, can effectively treat these symptoms. 

Berezhnoi et al. used a previously developed motion sequencing platform to evaluate movements of different mouse models for PD during early stages of disease pathology and during Levodopa treatment. The main advantage of this machine learning platform is that it can automatically detect subsecond postural changes from three-dimensional videos of animals. The researchers discovered that quicker, higher velocity movements were the first affected behaviors in early stages of PD. Levodopa improved movement speed at fine time scales but did not improve other attributes of these movements. 

Speaking on the implications of this work, Berezhnoi says, "Maybe applying the same machine learning approach we used could help identify early biomarkers for Parkinson's in people."